Eric Neumayer and Thomas Plümper Foreign terror on Americans Article (Accepted version) (Refereed)

Original citation: Neumayer, Eric and Plümper, Thomas (2011) Foreign terror on Americans. Journal of peace research, 48 (1). pp. 3-17. ISSN 1460-3578 DOI: 10.1177/0022343310390147 © 2011 The Authors This version available at: http://eprints.lse.ac.uk/30794/ Available in LSE Research Online: August 2012 LSE has developed LSE Research Online so that users may access research output of the School. Copyright © and Moral Rights for the papers on this site are retained by the individual authors and/or other copyright owners. Users may download and/or print one copy of any article(s) in LSE Research Online to facilitate their private study or for non-commercial research. You may not engage in further distribution of the material or use it for any profit-making activities or any commercial gain. You may freely distribute the URL (http://eprints.lse.ac.uk) of the LSE Research Online website. This document is the author’s final manuscript accepted version of the journal article, incorporating any revisions agreed during the peer review process. Some differences between this version and the published version may remain. You are advised to consult the publisher’s version if you wish to cite from it.

Electronic copy available at: http://ssrn.com/abstract=986191

0

Foreign Terror on Americans

Published in

Journal of Peace Research, 48 (1), 2011, pp. 3-17

Eric Neumayera and Thomas Plümperb

a Department of Geography and Environment, London School of Economics, London WC2A 2AE, UK and Centre for the Study of Civil War, International Peace Research Institute Oslo (PRIO) b Department of Government, University of Essex, Wivenhoe Park, Colchester CO4 3SQ, UK, and Centre for the Study of Civil War, International Peace Research Institute Oslo (PRIO)

Correspondence: e.neumayer@lse.ac.uk. Equal authorship. We thank three anonymous referees for their comments. Data and do-files to generate all the results reported and referred to in this article with Stata 10 can be found at http://www2.lse.ac.uk/geographyAndEnvironment/whosWho/profiles/neumayer/Home.aspx.

Electronic copy available at: http://ssrn.com/abstract=986191

1

Abstract

Americans are a major target of international terrorism. Yet, terrorists from some

countries are much more likely to attack American citizens than terrorists from other

countries. Similarly, anti-American terrorism from a specific foreign country is much

more prevalent during certain periods than others. This article develops a rational theory

of international terrorism, which argues that attacking foreign nationals is of strategic

value to terrorists even if they ultimately aim at gaining political influence in their home

country. Attacking foreigners is the more attractive to domestic terrorists the more the

terrorists’ home government depends on military support from the foreign country.

Applied to the US case, our theory predicts that more anti-American terrorism emanates

from countries that receive more US military aid and arms transfers and in which more

American military personnel are stationed, all relative to the country’s own military

capacity. Estimations from a directed country dyad sample over the period 1978 to 2005

support the predictions of our theory for both terrorist incidents involving Americans

and terrorist killings of Americans as dependent variables. These results are robust to a

wide range of changes to the empirical research design.

2

1. Introduction

Most terrorism is domestic and, not surprisingly, the victims of most attacks share their

nationality with the terrorists. However, US citizens are major victims of international

terrorism. Even before 09/11, at least 500 US citizens died in terrorist attacks conducted

by non-US terrorists from at least 37 different countries. In this article we develop and

empirically test the hypothesis that one important cause of anti-American terrorism is

US military support to foreign countries whose governments are in conflict with

terrorists. We augment existing rational explanations of terrorism and develop a theory

of international terrorism in general and of foreign terror on Americans in particular,

clearly specifying the conditions under which one would expect more anti-American

terrorism to emanate from some countries than others.1

In line with rational explanations of terrorism (Crenshaw, 1981, 2001; Pape,

2003, 2005; Kydd & Walter, 2006), we assume that terror groups want to gain a

significant political influence on their country of origin or the broader region.

Accordingly, terror groups are typically conflict with the government of their home

country. They attack foreign targets if they gain a strategic advantage or peer support

from doing so. We explain the strategic logic of attacking foreigners by the degree to

which foreign powers militarily support the government in the terrorists’ home country.

The US is the world’s leading arms exporter, provider of military aid and it

maintains military personnel in numerous countries around the globe. Yet, at the same

1 Our theory is of course not exhaustive. We do not claim that it can explain all

international or anti-American terrorism. We use the term “anti-American terrorism” as a

shortcut for terrorist attacks on Americans.

3

time it is also highly selective in who receives military support and how much. We

argue that more anti-American terrorism comes from countries whose governments

depend more on US military support. Accordingly, our theory predicts that more

terrorist attacks are plotted against Americans from countries, which receive more US

military aid and weapons and host more military personnel, all relative to their own

military strength.

We test this hypothesis with a directed dyadic dataset of American deaths caused

by foreign terrorists as well as terrorist incidents involving Americans, covering the

period between 1978 and 2005 using dyadic information which identifies the origins of

anti-American terrorism. To our knowledge, we are the first to develop a research

design that goes beyond an aggregate time-series analysis of anti-American terrorism

and which allows us to address both the temporal and the much larger spatial variation

in attacks on American citizens. Empirical findings provide ample support for the

hypothesis derived from our theory.

2. Rational Explanations of International Terrorism

Most rational theories of terrorism assume that terror groups do not regard terrorist

attacks as an end in itself. Rather, terrorist attacks are instruments used by (radical)

political groups which pursue goals and follow strategies (Fromkin, 1975; Crenshaw,

1981, 2001; Pape, 2003, 2005). Five goals have been associated with terror groups of all

kinds by Kydd & Walter (2006: 52): regime change, territorial change, policy change,

social control, and status quo maintenance. Without loss of generality, one can

summarize these goals under one umbrella notion: terror groups ultimately strive for

gaining political influence, power and control in their home country or wider region.

4

But, of course, not all extremist groups with such goals conduct terror acts. Why do

terror groups engage in terror rather than peaceful political action? The reason is that the

terrorist home country does not allow peaceful political participation (e.g. because it is

autocratic) or the goals of the terror groups are too unpopular to gain much support from

the population, or both (Enders and Sandler, 2006b). Either way, non-violent political

change is not an option.

Terror groups thus clash with the government of their home country. But it is an

asymmetric conflict, in which the weaker conflict party, the radical group, uses the age-

old weapon of the weak and powerless, terrorism to exert pressure on the stronger

conflict party (Pape, 2005). Taken together, only if three conditions simultaneously

apply, will political groups be likely to use terrorist means: first, the group has radical

political goals, which, second, puts it in conflict with the government and, third, the

group cannot achieve its goals by either legal means or by an open armed challenge on

the government. Terrorism, then, can be understood as the attempt of a weaker party in

a conflict over political goals to coerce the stronger conflict party to succumb to its

goals. If, in comparison, the government and the radical group have similar policy

preferences, no conflict emerges; if they have conflicting goals, but similar strengths,

then the leaders of an anti-government group are likely to form a guerilla group and can

directly confront the government in a civil war.

This rational logic behind terrorism does not explain, however, why terror

groups attack targets from other countries in general and from the US in particular. Yet,

three main factors explain why terrorists can benefit from attacking Americans:

opportunity, necessity and strategic value. The omni-presence of Western tourists,

business people, diplomats as well as commercial outlets across the world renders

5

attacking Westerners relatively easy and no other country is as prominently represented

in foreign countries as America (Pillar, 2001). Li & Schaub (2004) argue that economic

globalization tremendously increases such opportunity.

Sobek & Braithwaite (2005) develop the necessity argument for anti-American

terrorism. They state that if radical political groups – ‘revisionist actors’ as they call

them – want to challenge the status quo then they have to confront American dominance

in the world. Given their relative weakness, they need to resort to terrorism, the weapon

of the weak. Greater American dominance in the world means that the revisionist actors

‘experience a limitation in their ability to effectively alter the status quo’, which ‘will

thus likely lead them to more often choose terrorism over alternative actions’ (Sobek &

Braithwaite, 2005: 139f.). Terrorists then target America and Americans because the US

has the greatest interest and is most active in maintaining the status quo.

The strategic value of attacking Americans is highlighted by Crenshaw (2001) as

an answer to ‘Why America?’ In her view, not the extent of American dominance in the

world as such matters, but the specific decisions the US makes in choosing and

supporting allies. She argues that the US ‘has been susceptible to international terrorism

primarily because of its engagement on the world scene and its choice of allies.

Extremist groups in countries around the world have targeted United States interests in

an effort to achieve radical political change at home’ (Crenshaw, 2001: 432). Attacking

Americans is of strategic value to terrorists because they hope that America will retreat

if hit hard enough, which would facilitate the terrorists’ political struggle at home.

In sum, rational explanations regard anti-American terrorism as the consequence

of easy opportunity, as a political necessity due to the asymmetry in power between

weak terrorist groups and the most powerful country in the world, or as a strategic

6

choice due to the value that attacking Americans has for the terrorists’ campaign at

home.

Yet, not all rational explanations equally account for why America is targeted

more than other (Western) countries and why terrorists from certain countries attack

Americans much more than terrorists from other countries. The opportunity argument

fails to explain why other Westerners, which are after all not much less present in for-

eign countries than Americans, become significantly less victimized by foreign

terrorists even if we correct for population size. This argument also provides no

satisfactory motivation for terrorists to kill foreigners. Westerners and Americans are

not just attacked because of their mere presence and availability. Neither do Americans

become the victims of terrorists simply out of necessity. Whilst America is far more

powerful than any radical political group, so are many other countries in the world.

Thus, asymmetry in power constitutes a necessary but not a sufficient condition for

terrorism. A rational theory of anti-American terrorism needs to explain how and why

terrorists gain strategically from attacking US citizens and why terrorists of some

countries gain more so than terrorists from other countries.

In the remainder of this article, we will develop the idea of Crenshaw (2001) and

others further and we will argue that the strategic value to terrorists from attacking

Americans is a function of the extent to which foreign governments depend on

American military support. The greater American military support relative to domestic

military capacity, the more America represents an effective brake on the terrorists’ goal

of achieving political influence, power and control in their home country and the higher

the strategic value from attacking Americans.

7

3. A Strategic Theory of International Terrorism

In this section, we develop a theory of international terrorism, which we then apply to

the US case. The main purpose of the first step is to explain the benefit domestic

terrorists derive from attacking foreign targets. In the second step, we show how the

strategic logic of international terrorism renders radical groups from some countries

rather than others more likely to target Americans.

3.1. The Strategic Interests of Terrorists

Our theory firmly roots within the rational explanation of international terrorism. We

thus assume that terrorism serves as an instrument used by radical political groups

which aim at gaining influence on important policies and political control in their home

country (or wider region). We will show that under certain identifiable conditions

attacking foreign targets is in the strategic interest of the terror group.

Three distinctions are important for the development of our argument. First, we

distinguish between three types of actors: the terrorist group, the government of the

terrorists’ home country, with which the group is in conflict, and the foreign govern-

ment, which interferes in the domestic conflict on the side of the home government.

This distinction allows us to ultimately relate terrorist attacks on foreign citizens to a

political conflict between the terror group and its government – a conflict in which the

foreign power takes side. Accordingly, we argue that international terrorism emanates

from the foreign country’s influence on the domestic politics in the terrorists’ home

country – and not from a given conflict between the terror group and the foreign nation

per se.

8

Second, we distinguish between the leaders of the terror organization, the terror

entrepreneurs (Neumayer and Plümper, 2008), and their footmen, the terror agents.

Terror entrepreneurs have a crucial, decisive position in the terror organization. They

are the group’s leaders and behave predominantly strategically. From their perspective,

terrorism is a rational strategy in a political struggle to achieve an end, not an end in

itself. Terror agents, on the other hand, are attracted by the entrepreneurs’ ideology, but

do not hold a decisive or crucial position within the terror organization. Rather, they

serve as footmen of the entrepreneurs and follow specified or unspecified commands,

which may include the command of self-sacrifice in a suicide terror mission. Contrary

to terror entrepreneurs, terror agents need not behave predominantly strategically or

rationally. They may be driven by several, complex and varied motivations that may

also depend on the contextual and sometimes very personal circumstances (Victoroff,

2005). This second distinction thus allows us to leave many discussions in the literature

on terrorism aside – for example, the one on how ‘rational’ suicide terrorists are (Fer-

rero, 2006). A rational theory of terrorism focuses on the terror entrepreneurs, not their

agents, since they, not the agents, decide when, where and against whom terror attacks

are executed.

Our third and final crucial distinction is that between the ultimate goals of terror

groups on the one hand and intermediate strategic goals on the other hand. We argue

that the ultimate goal of terror entrepreneurs is to gain political influence, power and

control in their own country or wider region. Yet, most terror groups cannot reach their

ultimate goals within the foreseeable future. For this reason, they aim at achieving

intermediate strategic goals. In one way or the other, these goals all entail either a

9

weakening of the government, against which the terrorists fight, or a strengthening of

their own organization.

Terror groups can win a strategic advantage over the government in numerous

ways. For example, the terrorists can kill important government representatives. Killing

innocent individuals and bystanders can be equally beneficial to the terrorists, however,

if this raises sufficient media attention and causes the government to respond heavy-

handedly. The media attention becomes important since terrorists need access to a large

audience in order to increase support by like-minded individuals and recruit new agents.

A harsh government response has an ambiguous effect on the strength of the terror

organization since governmental crackdowns impose costs on the group. At the same

time, however, such crackdowns can also be beneficial to the terrorists if they are

perceived as over-reactions and lead to the restriction of individual freedoms and

increased costs for all citizens or particular groups of the population, which lowers the

government’s popularity. More importantly, perhaps, a harsh anti-terror response by the

government may infuriate the terror group’s actual and potential supporters, which

drives out more moderate voices and helps recruiting terror agents (Rosendorff &

Sandler, 2004; Richardson, 2006).2 It also facilitates gaining shelter and gathering

resources from those willing to support the terror group.

Yet, this explanation of the strategic interests of terror organizations does not

account for why these groups at times choose to attack foreigners. Since terror

entrepreneurs ultimately aim at changing the political system or important policies of

2 See Rosendorff & Sandler (2004); Bueno de Mesquita (2005a, b, 2007), Bueno de

Mesquita & Dickson (2007) and Powell (2007) for formal modeling of governmental

counter-terrorism.

10

their home country, one may even wonder why terrorists attack foreigners at all. It is to

the strategic value of attacking foreign targets to which we now turn.

3.2. The Strategic Value of Attacking Foreign Targets

Our micro-foundation of terrorism provides important insights to the logic of

international terrorism, suggesting at least three reasons why attacking foreigners may

become a utility-maximizing strategy for terror organizations. First, the nationality of

terror victims has a major impact on the extent of media attention. With global news

networks dominated by Western countries, it is attractive for any terror entrepreneur

anywhere in the world to inflict terror on nationals of Western countries as this is a sure

way of getting into global news. The media attention enables the terrorists to spread

their ideology more easily. Thus, a successful attack on foreigners from some countries

has a higher strategic value for terror entrepreneurs than a similarly successful attack on

foreigners from another country or on domestic citizens.

Second, attacking foreigners rather than domestic citizens may result in higher

peer acknowledgement for the terror group. This will be particularly the case if the

foreign citizens come from countries that are hated by the terrorists’ potential

supporters. Attacking civilians is difficult to justify in the eyes of most people, even

those sympathetic to the terror entrepreneurs’ goals. Yet, nationals from detested

foreign countries can be more easily de-humanized as infidels, suppressors, exploiters

and the like.

Third, and most importantly, some governments owe their survival partly to the

support of foreign countries, which may heavily influence the government’s policies

and may induce it to take immensely unpopular measures, heightening the conflict with

the domestic terrorists. Accordingly, the probability of terror groups reaching their

11

ultimate goal may depend on the amount of political and especially military support the

domestic government receives from abroad. Foreign powers can, for example, station

troops in the terrorists’ home country, provide military aid, arms, military training,

political and economic support. Thus, if foreign powers make it less likely for the terror

group to achieve its goals, an attack on foreign citizens can maximize the strategic

advantage of the terror group. By directly targeting the foreign supporter of the home

government, the terror entrepreneurs hope to trigger a decline in foreign support for the

home government or even its complete withdrawal.

Of the diverse forms in which foreign powers can bolster embattled gov-

ernments, military support is the one most relevant and most visible to the terrorists.3

Military aid, arms and training directly strengthen the government side in its struggle

with the terrorists, putting the terrorists at a disadvantage compared to the situation

without foreign military support. Military personnel of the foreign power may actively

engage in counter-terrorist actions. Even where they do not, they help the government

by releasing its own military personnel from some of its duties, thus lowering the

opportunity costs of counter-terrorist action.

3 This is not to say that political and economic foreign support may not also play a role in

international terrorism. We concentrate here on military support as the form of assistance

directly relevant to the terrorists’ violent struggle with the domestic government, leaving

an analysis of other forms of support by the foreign power to future research. It is also

worth noting that military aid is probably less fungible than general economic aid.

12

3.3. US Military Support, Anti-American Terrorism and the Terrorists’ Countries of

Origin

The logic of international terrorism discussed above can be applied straightforwardly to

make predictions on the countries of origin from which terrorists come who conduct

terror attacks on US targets. Our theory posits that anti-American terrorism more likely

originates from countries in which US military support has the largest effect on the

military capacity of the government receiving this support.

More governments borrow their military strength from the US than from any

other country. Yet, the US is also highly selective in deciding which foreign

governments receive military support. It supports some countries but not others. In turn,

some countries and their governments depend much more on US military support than

others.

Military support can take many different forms. Military aid and arms exports

buttress the military capacity of a country’s government. While Israel, Egypt, Turkey,

Greece and Pakistan are among the largest recipient countries of US military aid during

our period of study in both absolute and relative terms, several Central American

beneficiaries also rank prominently among the countries receiving a significant US

contribution relative to their total military expenditures. The North Atlantic Treaty

Organization (NATO) partners, Japan, Egypt, Israel, and Saudi-Arabia receive the

largest amount of US arms exports. If we standardize arms imports by a country’s

domestic military expenditures, however, we find that US arms transfers are most

important for some Arab countries like Egypt, Bahrain and Jordan. Another form of

military support to foreign governments occurs when the US stations its own military

personnel in the country. US military bases are spread all over the world. In absolute

13

numbers most US overseas troops are stationed in some of the NATO countries, South

Korea, Japan, and the Philippines, which reflects the past American geo-political

interest in containing communism. If we disregard NATO countries and the troops

currently stationed in Iraq and Afghanistan, then Panama, Bahrain, Kuwait, and Hondu-

ras host the largest ratio of US troops relative to their own troops during our study

period.

The flipside of this military support for foreign countries is that the more im-

portant US support becomes for the stability of the terrorists’ home government, the

more attractive to terrorists become attacks on US targets. In countries largely

dependent on US military support, not only the military strength of the home

government frustrates the terrorists’ bid for political influence, power and control, but

also the extent to which the US allows the home government to borrow from its much

larger military strength.

In targeting Americans, the terrorists attempt to drive a wedge between the home

government and its foreign supporter. Thereby, they seek to increase the costs for the

US beyond the gains the US derives from supporting the government. If they manage to

achieve this goal, the terror attacks may ultimately cause a withdrawal of US troops and

a significant reduction in US arms exports and military aid. This would destabilize the

government the US had supported.

The terrorists derive succor from, admittedly highly selective, past experience.

The suicide attacks on US marine barracks in October 1983, for example, which killed

14

241 Americans, made the US withdraw its troops from Lebanon.4 Osama bin Laden has

repeatedly professed his conviction that the US and its soldiers are cowards who will

give in if only hit hard and long enough (Al Qaeda, 2006a, b). Rightly or wrongly, Al

Qaeda leaders believe that their terror campaigns will ultimately induce the US to

withdraw from the Middle East and that this would enable Al Qaeda and like-minded

groups to expand their political influence and, perhaps, even allow them to take control

of pivotal countries like Egypt and Saudi Arabia (Posen, 2001/02).

In sum, our theory predicts that groups from countries whose government de-

pends on US military support are most likely to conduct terror on Americans. Our

testable hypothesis is that a foreign government’s greater dependence on US military

aid, arms exports, and military personnel significantly increases attacks by terrorists of

this foreign country on US citizens. In the remainder, we put this hypothesis to an

empirical test, starting with a discussion of the research design.

4. Research Design

Empirically, social scientists can address the special case of anti-American terrorism

from two different perspectives: First, time-series analyses of terrorist attacks on US

targets identify the peaks in the number of terrorist attacks on Americans and their

correlates. For example, Sobek and Braithwaite (2005) find that the total aggregate sum

of all terror attacks against US targets is a function of overall American political and

4 This was openly admitted to by the late President Ronald Reagan in his memoirs: ‘The

price we had to pay in Beirut was so great, the tragedy at the barracks was so enormous…

We had to pull out.’ (cited in Pape, 2005: 55).

15

military dominance in the world. Second, a (pooled) time-series cross-sectional dyadic

analysis with terrorist attacks on Americans as the dependent variable identifies the

conditions determining the large variation in anti-American terrorism both across time

and, more importantly, across terrorists’ countries of origin. It is only this latter

perspective, which we adopt here, which sheds light on the causes of anti-American

terrorism. This is because while such terrorism is widespread across the world, there are

also clear and stark differences in its extent across terrorists’ countries of origin. In fact,

any analysis of anti-American terrorism needs to tackle the fact that, even excluding the

9/11 attacks, terrorists from only ten countries account for roughly 80 per cent of all

American casualties and that the top ten countries of anti-American terrorism in terms

of terrorist incidents account for roughly half of all terrorist acts.5

4.1 The Dependent Variables

Terrorism is notoriously difficult to measure because clear-cut definitions that allow

distinguishing terrorism from guerrilla warfare on the one hand (Sambanis 2008) and

terrorism from ordinary crime on the other do not exist. We use data from the

‘International Terrorism: Attributes of Terrorist Events’ (Iterate) dataset (Mickolus et

al., 2003), which defines terrorism as ‘the use, or threat of use, of anxiety-inducing,

extra-normal violence for political purposes by any individual or group, whether acting

for or in opposition to established governmental authority, when such action is intended

5 According to our data, the top 10 countries in terms of anti-American terrorist incidents

are Colombia, the Philippines, Greece, Turkey, El Salvador, Lebanon, Peru, Iran,

Germany and Pakistan. In terms of terrorist killings of US citizens, Saudi Arabia, Iraq,

Egypt and Jordan enter the group, while Greece, Peru, Iran and Germany drop out.

16

to influence the attitudes and behavior of a target group wider than the immediate

victims’ (ibid.: 2). Terrorist violence includes as diverse incidents as, among others, as-

sassinations, bombings and armed attacks, arson and fire, kidnapping, and skyjacking.

However, ordinary crime and violence for purposes other than political are explicitly

excluded. Violence committed during international and civil wars is not coded as

terrorism either. Consequently, guerrilla attacks by rebel groups remain uncounted,

unless they target civilians or the dependents of military personnel (Mickolus, Sandler

& Murdock, 1989: xii). Thus, Iterate excludes terror attacks against US soldiers in Iraq,

but includes attacks against American civil workers (Enders & Sandler, 2006a: 372).

Iterate provides information on whether US citizens have been victimized or

even killed in terror attacks. To create a directed country dyadic dependent variable we

use information on the nationality of the terrorists (the first nationality of terrorists in

case more than one nationality is involved). We employ two dependent variables. One is

the annual sum of US citizens killed by terrorists, the other one the annual sum of all

terrorist incidents that victimized US citizens. The location of the terror incident as such

does not matter. Thus, terrorists from a certain country might inflict terror on US

citizens in their own country, in a third country or even in America itself.6

Anti-American terrorist incidents emanate from terrorists with nationality from

91 different countries during the period of our study (1978 to 2005), but as pointed out

already, we observe substantial variation in the amount of terrorism across terrorists’

home countries among these 91 countries. Terrorists from 39 different countries killed

6 The latter is a rather rare thing to happen, however, despite the 9/11 attacks. Enders &

Sandler (2006a) find evidence for a regional shift of anti-American terror attacks from the

Western hemisphere and Africa to the Middle East and Asia after 9/11.

17

568 Americans, not counting those killed in the 9/11 attacks other than the ones inside

the airplanes. It is important to note, however, that despite the generally high levels of

public concern and despite the fact that Americans are major victims of international

terrorism, foreign terror on Americans remains a relatively rare event. Terrorist

incidents involving Americans occur in only about 7.8% of dyad years and Americans

get killed in only about 1.8% of dyad years.

4.2. Explanatory Variables

We use three different variables to capture the various ways in which the US chooses to

support foreign governments militarily. Our main explanatory variables are thus US

arms exports, military aid, and military personnel stationed abroad. Data on arms

exports have been provided to us, courtesy of the Stockholm International Peace

Research Institute (SIPRI). Data on military aid and military personnel come from

USAID (2006) and US Department of Defense (various years), respectively. Military

aid covers both military grants and loans, principally for the acquisition of US defense

equipment, services, and military training. Active duty military personnel come from all

services (army, navy, marine corps and air force). The three variables do not always

move in the same direction. Some recipient countries receive a large amount of military

support in one or two dimensions, but not necessarily in the remaining ones.

In order to measure the relative dependence of the terrorists’ home country on

US military support, we have to relate US arms exports, aid and military personnel to

some measure of domestic military strength. For arms exports and aid we divide their

total value by domestic military expenditures. For military personnel, we compute the

18

ratio of US to domestic military personnel.7 Data on domestic military spending and

personnel come from the Correlate of War’s Composite Index of National Capacity

(CINC) measure as well as from World Bank (2006).8

As control variables, we include the log of per capita income, the level of de-

mocracy, the logged population size of the terrorists’ home country and the

geographical distance between it and the US. This follows arguments provided by

Krueger & Laitin (2008) and Abadie (2006) on welfare and terrorism and Enders &

Sandler (2006b) on regime type and terrorism. We control for population size to

account for the simple fact that, ceteris paribus, more populous countries may generate

more anti-American terrorism. Lastly, we account for the fact that geographical pro-

ximity may lower the costs for terrorists to target Americans. The World Bank (2006)

provides data on income and population. The democracy variable is measured by the

Polity project’s polity2 variable (Marshall, Jaggers & Gurr 2006), which runs from -10

to 10. The natural log of the distance between Washington D.C. and the capital city of

the terrorists’ home country stems from Bennett & Stam (2005).9

7 One referee suggested using GDP and population as weighting variables instead. We did

not follow this advice since we regard our variables as better proxies for a country’s

capacity to deal with domestic terrorist threats. In any case, military expenditures and

military personnel are highly correlated with GDP and population, respectively.

8 We mainly use CINC data and World Bank data only as complement. In the period of

overlap, the two data sources are so very highly correlated with each other (r = .97) that

pooling them is no problem.

9 For lack of data, we cannot control for the amount and quality (Bueno de Mesquita, 2007)

of governmental counter-terrorist measures.

19

4.3. Estimation Method and Models

Our dependent variables are count variables (number of US citizens killed in terrorist

attacks and number of terror incidents involving Americans). For all reported results,

the negative binomial estimate is more reliable than the Poisson model, because the

sample variance of the number of killings and incidents exceeds its sample mean by a

factor of approximately 35 and 7, respectively. In robustness tests, we also used a

variant of the negative binomial called the zero-inflated negative binomial. We compute

standard errors adjusted for clustering on the terrorists’ home country, though the varia-

tions in killings and incidents over time are large and clustering is therefore of minor

importance. Our sample covers the period 1978 to 2005 and up to 149 countries.10 Due

to missing data on the explanatory variables not all possible observations can be

included in the analysis. We do not include year dummies to account for trends in

foreign terror on Americans, but our results remain fully robust if we do.

5. Empirical Evidence and Analysis

Our theory predicts that radical terrorist groups are more likely to target US citizens

when United States’ military support stabilizes the government in the home country of

the terrorists. In this section we test our hypotheses. Before we present the results of our

regression analyses, however, we briefly discuss some illustrative evidence.

El Salvador provides a good example for illustrating the link between US

military support on the one hand and anti-American terrorism on the other. With the

10 The Iterate data go back to 1968, but the project starts reporting separately on US victims

only from 1978 onwards.

20

start of the civil war in 1980, the US quickly stepped up its military aid to the El

Salvadoran government from almost nothing to very high levels throughout the 1980s,

reaching on average about 50% of domestic military expenditures. As a result, there was

a wave of terrorist attacks on Americans throughout the 1980s (40 incidents in total

between 1980 and 1991), whereas there had been practically no terror attacks on

Americans before. As the US military aid waned to very low levels in the beginning of

the 1990s, anti-American terrorism virtually disappeared, with only two recorded

incidents since 1992.

Similarly, in Guatemala no anti-American terrorism was recorded in the 1970s.

When the US sent large amounts of arms to the Guatemalan government in the early

1980s, a wave of terror attacks on Americans ensued, lasting for about half a decade.

Likewise, a period of no anti-American terrorism since 1968 was followed by a range of

attacks in the early 1980s and then again in the late 1980s when the US sent massive

amounts of military aid and large amounts of arms to the government of Honduras.

As pointed out already, US military support for foreign governments can take

several forms, which need not move together. In the case of El Salvador and Guatemala,

the US had very few military personnel stationed in the countries. In Honduras, there

was at times a larger contingent of US troops stationed, but the number fluctuated over

the years. The Philippines provides a better example of how a large contingent of US

military personnel makes Americans vulnerable to terrorist attacks by domestic

terrorists. Before the US reduced its personnel stationed in the Philippines from around

15,000 (equivalent to about 12% of the domestic armed forces) to below 300 from 1993

onwards in tandem with reducing its military aid from about 10% relative to domestic

military expenditures to much lower levels, US citizens suffered a total of 59 terrorist

21

attacks between 1978 and 1992. In contrast, there were only 20 anti-American terrorist

attacks from 1993 to 2005 despite the continuing and ongoing civil conflict between the

Philippine government and rebel groups.

A more recent example is Saudi Arabia. In the wake of the invasion of Kuwait

by Iraqi forces and the ensuing Gulf War, the US temporarily stationed a large number

of troops in the country. These were soon afterwards decreased, but the US continued to

deliver large amounts of weapons to the Saudi regime throughout the 1990s. From 1995

to 2000, 43 Americans in total were killed by Saudi terrorists, making Saudi Arabia one

of the top countries of anti-American terrorist origin even before the 9/11 attacks. There

were no recorded attacks by Saudi terrorists on Americans before 1995.

This pattern between US military support and anti-American terrorism is not

confined to these selected country cases, but can be observed more generally. Whereas

the group of top 20 countries in terms of killings of US citizens have a mean ratio of US

military troops relative to domestic troops of just below 3%, the remaining countries

have an average of just 1.2%. The top 20 receive an average of 8.2% of US military aid

and 5.8% of US arms exports relative to domestic military expenditures, the rest only

2.5% and 2.2%, respectively. A similar pattern emerges for the group of top 20

countries in terms of terrorist incidents.

The illustrative evidence presented above is indicative and consistent with our

argument, but this brief discussion does not provide a sufficient test of the hypotheses

derived from our theory. We thus move to a multivariate statistical analysis.

5.1 Analysis

We can look at our empirical estimation problem from two perspectives. On the one

hand, US arms exports, military aid, and military personnel stationed abroad can have

22

an independent and additive effect on the supported government’s ability to survive the

political challenge by a radical group. On the other hand, however, one could argue that

we are interested in an unobservable latent variable, which can be approximated by

these three observable variables. A simple cross-correlation may help to decide whether

our three military support variables can be unified to a sensible ‘latent’ dimension.

Table I shows the cross-correlation of US military aid, arms export and military

personnel.

Table I: Cross-Correlation of dependence on US military aid, personnel and arms

exports, 1978-2005

military aid arms exports military per-sonnel

military aid dependence 1.00 arms export dependence .14 1.00 military personnel dependence .00 .02 1.00

Table I reveals that for the United States, military aid, arms exports and military

personnel appear to be substitutes rather than complements. The partial correlation

coefficients remain low if we exclude relatively wealthy countries (which do not need

military aid) from the analysis, even though the correlation between military aid and

arms exports increases slightly.

Accordingly, this simple analysis suggests that our three military support

variables cannot usefully be combined into their principal components or a single factor

score. In turn, we may use these three variables simultaneously in one regression

without having to fear that the co-variation leads to inefficient and thus unreliable

estimates (Plümper & Troeger, 2007). This interpretation finds support by a principal

component analysis, which suggests the existence of two principal components (an ‘aid

23

– arms exports’ dimension and a military personnel dimension). Similarly, a maximum

likelihood factor analysis shows high values of ‘uniqueness’, suggesting that the factor

score inappropriately describes the original variables.

We therefore conduct only the first type of analysis, where we use the various

military assistance variables as independent and additive determinants of terrorist

activities on Americans. Table II displays the results of models, in which we first

include one ‘military support’ variable at a time for the number of US citizens killed as

dependent variable. The exclusion of other military support variables is unlikely to bias

the estimates due to the low correlation between these variables. Nevertheless, we also

show a model with all military support variables entered simultaneously. Table II: Negative binomial estimates of number of US citizens killed in terrorist attacks model k1 model k2 model k3 model k4 ln population of terrorists’ home country

0.707 (0.152) ***

0.612 (0.146) ***

0.645 (0.155) ***

0.746 (0.153) ***

ln distance between terrorists’ home country and US

0.098 (0.135)

0.028 (0.225)

-0.034 (0.361)

0.126 (0.119)

ln per capita income of terrorists’ home country

0.177 (0.178)

0.064 (0.164)

0.112 (0.167)

0.102 (0.184)

level of democracy in terrorists’ home country

-0.078 (0.035) *

-0.061 (0.033) *

-0.071 (0.031) *

-0.087 (0.035) *

terrorists’ home country dependence on US military aid

0.055 (0.019) **

0.049 (0.020) *

terrorists’ home country dependence on US arms exports

0.078 (0.028) **

0.039 (0.021) *

terrorists’ home country dependence on US military personnel

0.029 (0009) ***

0.034 (0.006) ***

intercept -16.811 (3.961) ***

-13.810 (4.151) **

-13.962 (5.222) **

-17.347 (4.073) ***

ln alpha 3.561 (0.279) ***

3.657 (0.270) ***

3.695 (0.287) ***

3.540 (0.273) ***

N obs. 3360 3360 3483 3341 Wald chi² 30.32 *** 38.80 *** 25.32 *** 60.97 *** -log likelihood 549.05 554.45 563.96 539.39 Standard errors clustered on terrorists’ home country in parentheses, * p(z)<0.1 ** p(z)<0.01 *** p(z)<0.001.

24

Before we come to our variables of main interest and thus to our hypothesis, we briefly

discuss results on the control variables. More populous countries account for more anti-

American terrorism, all other things equal, as expected. The US seems to be an outlier

with respect to the impact of distance on terrorism. In a much larger sample including

all country dyads for which data was available, we found a highly significantly negative

effect of distance on terror incidents and victims (Neumayer & Plümper, 2008; Plümper

& Neumayer, 2008). The global reach and projection of US power apparently

invalidates the law of distance that applies to terror attacks against other countries in the

world. Interestingly, in line with other studies we also find that per capita income in the

potential origin countries of terrorists has no statistically significant impact. This should

at least caution against suggestions that international terrorism will wane with economic

development in foreign poor countries – a position at times held by (among others)

former UN General Secretary Kofi Annan, and US Presidents Bill Clinton and George

W. Bush (see Piazza, 2006). Our results do provide some weak support, however, to

those who suggest that a higher level of democracy in foreign countries will lower anti-

American attacks (see Gause III, 2005).

Turning to our variables of main interest, we find evidence in favor of our

hypothesis. In particular, the three ‘military support’ variables exert a positive impact

on the number of US victims. All three show a significant and positive relation to the

number of American terror victims, as per our theory. The three variables also exert a

substantively important influence on the dependent variable. A one standard deviation

increase in the measure of military aid, arms exports and military personnel raises the

expected count of anti-American terrorism by 135, 109 and 24%, respectively.

25

Model 4 of Table II shows that the predictions of our theory still find support,

even if we include the military support simultaneously. An analysis of substantive

effects based on model 4 allows one to gauge the relative importance of the three

measures of military support when entered together in the estimations. A one standard

deviation increase in US military aid raises the expected count of anti-American

terrorism most by 114%, followed by arms exports and military personnel, of which a

one standard deviation increase leads to an increase of 45 and 30%, respectively. It

would thus appear that US military aid matters most for foreign terror on Americans.

In Table III we account for the fact that terrorists can attack and, in fact, wound

Americans without necessarily managing to kill them. It repeats the models from Table

II, but this time with terrorist incidents involving American citizens as the dependent

variable.

26

Table III: Negative binomial estimates of number of terrorist incidents with

American victims

model i1 model i2 model i3 model i4 ln population of terrorists’ home country

0.668 (0.130) ***

0.617 (0.139) ***

0.645 (0.143) ***

0.677 (0.131) ***

ln distance between terrorists’ home country and US

-0.103 (0.442)

-0.188 (0.586)

-0.215 (0.617)

-0.058 (0.343)

ln per capita income of terrorists’ home country

0.154 (0.103)

0.066 (0.108)

0.076 (0.113)

0.083 (0.111)

level of democracy in terrorists’ home country

-0.001 (0.023)

0.010 (0.023)

0.006 (0.023)

0.002 (0.024)

terrorists’ home country dependence on US military aid

0.044 (0.013) **

0.038 (0.014) **

terrorists’ home country dependence on US arms exports

0.044 (0.016) **

0.024 (0.016)

terrorists’ home country dependence on US military personnel

0.027 (0.009) **

0.027 (0.006) ***

intercept -12.766 (3.672) ***

-10.522 (4.598) *

-10.743 (4.907) *

-12.899 (3.095) ***

ln alpha 2.219 (0.131) ***

2.269 (0.128) ***

2.269 (0.134) ***

2.178 (0.136) ***

N obs. 3360 3360 3483 3341 Wald chi² 47.48 *** 40.91 *** 38.63 *** 84.79 *** -log likelihood 1765.98 1779.53 1802.06 1749.39 Standard errors clustered on terrorists’ home country in parentheses, * p(z)<0.1 ** p(z)<0.01 *** p(z)<0.001 , see Table II for the model notation

As can be seen, the results are rather similar. Most importantly, our hypothesis still

finds support. In the model with all support variables enclosed, the coefficient of the US

arms exports variable becomes marginally insignificant. One potential explanation for

this finding is that arms exports are the weakest proxy for military support. If the US

does not provide arms, the odds are that other countries will step in and sell them to a

government willing to pay. Thus, the entrepreneurs of terror have not as much to win

from attacking the provider of weapons to the government it is in conflict with,

compared to foreign providers of other forms of military support. Terror entrepreneurs

should respond most clearly to military aid and the presence of foreign troops. The

27

former often provides military training for the government’s troops, while the latter may

support the home government directly in case of a severe domestic threat.11 In terms of

substantive importance, the estimations on anti-American terror incidents confirm the

results from the estimations on terrorist killings of Americans: military aid has the

strongest impact followed by military personnel and arms exports, which is, however,

subject to the caveat regarding arms exports stated above.

5.2 Robustness

Our results are robust to a number of changes to the research design. To avoid multiple

counting, only the first nationality of the terrorists determines the origin country of a

terrorist act. This has the disadvantage that information on the second and third primary

nationality of terrorists, also coded in Iterate, is lost, but the vast majority of terrorist

acts only involve one nationality of terrorists. Our results are robust to attributing

terrorist acts to all the first three main nationalities of terrorists simultaneously. We

exclude the few cases for which Iterate does not provide information on the primary

nationality of terrorists or victims. Also, we do not include terrorist acts committed by

‘Indeterminate Arabs, Palestine’ in the estimations. Our results remain robust if we

allocate each of these terror attacks to a randomly drawn Arab country.

An additional problem occurs because in a small number of cases, Iterate indi-

cates that Americans were killed, but states the exact number as unknown. The 9/11

11 Another interesting difference is that democracy is no longer statistically significant for

incidents, whereas it was significant for killings. This could reflect the fact that Islamist

terrorists tend to come from autocratic countries and Islamist terrorism on average kills

more people per terrorist act than other terrorism, even without 9/11 (Rapoport, 2004;

Enders and Sandler, 2006b).

28

terror attacks are the most prominent case: 189 American victims are stated (those on

the planes), but for those killed on the ground, Iterate does not state a number, probably

because it remains unknown how many of the 2973 non-terrorists killed in total were

Americans (National Commission on Terrorist Attacks upon the United States, 2004:

552). Less clear is why Iterate attributes the attacks to the group of ‘Indeterminate

Arabs’, when 15 of the 19 hijackers were Saudi Arabian. Our results are robust toward

counting, as a conservative short cut, two thirds of victims as Americans and allocating

the attacks to Saudi Arabia as the origin country.

Anti-American terrorism is a relatively rare event and some countries never

produce any such terrorism. Our dependent variables have thus a large number of zeros

and some dyads only have zeros. We therefore also estimated the models with all the

three military support variables included simultaneously with a zero-inflated negative

binomial estimator. This is a maximum likelihood estimator, which assumes that some

observations take on a value of zero with probability of one (Long & Freese, 2006). The

zero-inflated model would thus be appropriate if there were a variable which determined

with certainty that no anti-American terror occurred for some cases. This is a

problematic assumption since we see no theoretical reason why a particular dyad year

has a probability of one of not producing any anti-American terror. Nevertheless, we

employed this estimator in robustness tests as it is often (if falsely) used in situations

where the share of zeros is deemed to be ‘large’. We found that the results from the

zero-inflated negative binomial estimator were consistent with their negative binomial

equivalent.12

12 In a zero-inflated negative binomial regression model, a variable can affect the odds of

always 0 and/or affect the expected count for those not always 0, as there are two

29

The US often renders military support to governments fighting a civil or

international war. If such armed conflict raises general levels of violence, then US

military support will probably be correlated with an elevated risk of terrorism, whilst

both are co-determined by armed conflict. This raises the question whether our results

spuriously pick up an effect of warfare. Yet, our results, while confirming that warfare

increases terrorism, are robust to including both small and large international and civil

armed conflicts, using data from the Uppsala/PRIO dataset (Gleditsch et al., 2002).

Anti-American terrorism does not always remain limited to terror groups that

oppose their domestic government. The US State Department also accuses several for-

eign governments of sponsoring terrorism. If we include a dummy variable, which is set

to one for years in which another country was accused of as a state sponsor of terrorism

(data from Office of the Coordinator for Counterterrorism, various years), results

remain substantively unchanged. The dummy variable itself is never statistically

significantly different from zero suggesting that state sponsorship of terrorism has no

significant effect on the number of Americans killed. This could be either because state

sponsorship of terrorism remains unimportant, the US allegations point into a false

direction or because these governments sponsor anti-American terrorism by nationals

other than their own. Unfortunately, due to the clandestine nature of such activities, we

estimating equations (inflation and count equation). For the sum of killings of Americans

as dependent variable, we found that military aid and military personnel raise the

expected count of killings for those that are not always zero. For the sum of anti-

American terror incidents as dependent variable, these military support variables were

marginally insignificant in the count equation, but they significantly lower the odds of

always zero in the inflation equation and thus raise the likelihood of terror incidents ever

occurring.

30

cannot possibly control for state sponsorship of terror by nationals other than the

sponsoring state.

Finally, one might also wonder whether our results suffer from reverse causality

bias. Perhaps, instead of raising the strategic benefits to foreign terrorists and thus

indirectly leading to anti-American terrorism, US military support goes to countries,

which are the major sources of anti-American terrorism, with the intention to boost their

counter-terrorism capacity. If such reverse causality exists, then it should bias the

coefficients of our military support variables upwards. We stress that US military

support going to foreign allies, which are faced by a high or increasing domestic

terrorist threat does not constitute reverse causality, but is fully consistent with our

theory. There is potential reverse causality if, and only if, more US military support

goes to countries from which high or increasing terror on Americans emanates. The

standard textbook approach to endogeneity recommends using instrumental variables

for the offending explanatory variables. As almost always, valid and efficient

instruments are very difficult to find. We tackled this potential problem with two

different approaches. First, we used a two-stage instrumental variable negative binomial

approach, in which we used the relevant US military support variables lagged by ten

years. This provides a good instrument for short-term or temporary increases in US

military support to countries, which have experienced major anti-American terrorism

because practically no anti-American terror campaign included in our dataset lasts

longer than a period of ten years. All the military support variables remain statistically

significant in the instrumental variable regressions, albeit at lower z-ratios. Second, we

used another approach that does not depend on instrumental variables, the validity of

which can always be contested. If reverse causality existed, then it should have become

31

much more pronounced after the attacks of 9/11 when the US stepped up military

support to countries suffering from terrorist threats, particularly threats to US citizens,

and willing to join the US in its global ‘war on terror’. As a consequence, the upward

bias in the estimated coefficients for the military support variables should be much more

pronounced in the period 2002 onwards. In order to conduct this indirect test of reverse

causality, we interacted all military support variables with a dummy variable for the

post-9/11 years (2002 to 2005). However, these interaction effects were never

statistically significant. Additionally, we tested whether it makes any difference to our

results if we restrict the sample to the period until 2001, thus excluding the years in

which potential reverse causality, if at all present, is likely to have been strongest.

Again, the results remain largely unaffected. Together, we take this as evidence that our

results are not spuriously driven by reverse causality bias.

5.3 Extensions

Of course, the logic of our strategic theory is not confined to American military support

for foreign governments and foreign terror on Americans. It extends to military support

from other countries to foreign governments and therefore to foreign terror on citizens

from these other supporting countries. We restricted our analysis here to the US case

because for the US data on military personnel stationed abroad and military aid given

are available. Other countries do not report such statistics, nor are reliable data readily

available from third parties. We have, however, data on bilateral arms exports from the

ten major arms exporters, taken from SIPRI (2007). If we estimate a model of foreign

terror on individuals from these major arms exporters (other than the US), then

consistent with our theory we find that a larger value of arms exports to domestic

32

military expenditures raises terror from the arms importing country on individuals from

the arms exporting country. Moreover, in Plümper & Neumayer (2008) we demonstrate

for a global sample that countries that support each other (as proxied by military

alliances), are more likely to encounter terrorist attacks from foreign terrorists of the

allied countries and the more so the stronger the foreign ally relative to the terrorists'

home country.

6. Conclusion

This article has argued that terrorist groups aim at gaining a significant influence on

their home country’s political system or policies and that, therefore, they have an

incentive to attack targets from foreign countries if the foreign government militarily

supports the domestic government of the terrorists’ home country. American military

support to foreign governments creates an incentive for foreign terrorist groups to attack

Americans.

Our study is, to our knowledge, the first one to analyze the link between

American foreign military policy and foreign terror on Americans quantitatively. Only

regression analyses in a sample with large spatio-temporal variation can identify

systematic patterns, even though this of course does not prove the causal mechanism to

be correct. Rather, we have empirically demonstrated that the more governments in

other countries are dependent on US support in terms of military aid, arms exports and

stationed military personnel the more attacks against Americans and the more killings

of Americans by terrorists from these countries can be expected. We therefore found no

reason for rejecting our theory.

33

Data constraints on forms of military support other than arms exports prevent us

from undertaking a similar study for other powers which militarily support foreign

governments. However, the results from our analysis as well as the policy implications

should, to mention but two examples, similarly apply to anti-Russian terror in response

to Russia’s support to Central Asian governments in conflict with domestic terrorists or

to French support for Algeria and Lebanon (see Shughart, 2006).

What are the policy implications of our analysis? Our results suggest that

Americans will, on average, be less at risk of terrorism if the US reduces or even

withdraws its military support from countries whose governments are heavily dependent

on US support. They do not support those who argue that ‘there is no reason to assume

that terrorist enemies would let America off the hook if it retreated’ (Betts, 2002: 34),

even though a reduction in anti-American terrorism may come with a significant time

delay after the US retreats. In the short run, terror organizations may even increase the

number and severity of attacks on US targets in order to demonstrate to their peers a

causal link between their attacks and US retreat. In the long run, however, we expect

terrorist activities against American targets to decrease, because the strategic benefit

from attacking Americans declines as America reduces its military support. Note,

however, that if terror groups such as Al Qaeda have wider regional ambitions then

withdrawing military support merely from one country in the region without reducing

support to other countries simultaneously is unlikely to reduce the terrorists’ incentive

to target Americans.

It is an entirely different question whether withdrawing military support to

countries in which governments face violent opposition from terror groups serves

America’s best interest. The benefit of reduced risk of terrorism needs to be carefully

34

balanced against the costs of exposing a weakened government to the terrorists’

challenge. The political, economic, and social gains from supporting political friends

may outweigh the costs of foreign terror on Americans.

35

References

Abadie, Alberto, 2006. ‘Poverty, political freedom, and the roots of terrorism’, Ameri-

can Economic Review 96(2): 50-56.

Al Qaeda, 2006a. ‘Declaration of War Against the Americans Occupying the Land of

the Two Holy Places’, in David C. Rapoport, ed., Terrorism – Critical Concepts

in Political Science. Volume IV. London and New York: Routledge (271-295).

Al Qaeda, 2006b. ‘Jihad Against Jews and Crusaders’, in David C. Rapoport, ed.,

Terrorism – Critical Concepts in Political Science. Volume IV. London and New

York: Routledge (295-297).

Bennett, D. Scott & Allan C. Stam. 2005. EUGene – Expected Utility Generation and

Data Management Program (http://eugenesoftware.org/).

Betts, Richard K., 2002. ‘The Soft Underbelly of American Primacy: Tactical Advan-

tages of Terror’, Political Research Quarterly 117(1): 19-36.

Bueno de Mesquita, Ethan, 2005a. ‘The Quality of Terror’, American Journal of Po-

litical Science 49(3): 515-530.

Bueno de Mesquita, Ethan, 2005b. ‘Conciliation, Counterterrorism, and Patterns of

Terrorist Violence’, International Organization 59: 145-176.

Bueno de Mesquita, Ethan, 2007. ‘Politics and the Suboptimal Provision of Counter-

terror’, International Organization 61(1): 9-36.

Bueno de Mesquita, Ethan & Eric S. Dickson, 2007. ‘The Propaganda of the Deed: Ter-

rorism, Counterterrorism, and Mobilization’. American Journal of Political

Science 51(2): 364-381.

Crenshaw, Martha, 1981. ‘The causes of terrorism’, Comparative Politics 13(4): 379-

399.

36

Crenshaw, Martha, 2001. ‘Why America? The Globalization of Civil War’ Current

History 100(650): 425-432.

Enders, Walter & Todd Sandler, 2006a. ‘Distribution of transnational terrorism among

countries by income class and geography after 9/11’, International Studies

Quarterly 50(2): 367-393.

Enders, Walter & Todd Sandler, 2006b. The Political Economy of Terrorism. Cam-

bridge: Cambridge University Press.

Ferrero, Mara, 2006. ‘Martyrdom contracts’, Journal of Conflict Resolution 50(6): 855-

877.

Fromkin, David, 1975. ‘The Strategy of Terrorism’, Foreign Affairs 53(4): 683-698.

Gause III, F. Gregory, 2005. ‘Can Democracy Stop Terrorism?’ Foreign Affairs 84(5):

62-76.

Gleditsch, Nils Petter, Peter Wallensteen, Mikael Eriksson, Margareta Sollenberg &

Håvard Strand, 2002. ‘Armed Conflict 1946–2001: A New Dataset’, Journal of

Peace Research 39(5): 615-37.

Krueger, Alan & David Laitin, 2008. ‘Kto Kogo? A Cross-Country Study of the Origins

and Targets of Terrorism’, in Phil Keefer & Norman Loayza, eds., Terrorism,

Economic Development, and Political Openness. New York: Cambridge

University Press (148-173).

Kydd, Andrew H. & Barbara F. Walter, 2006. ‘The strategies of terrorism’, Interna-

tional Security 31(1): 49-80.

Li, Quan & David Schaub, 2004. ‘Economic Globalization and Transnational Ter-

rorism’, Journal of Conflict Resolution 48(2): 230-258.

37

Long, J.Scott & Jeremy Freese, 2006. Regression Models for Categorical Dependent

Variables Using Stata. College Station, TX: Stata Press Publication.

Marshall, Monty G., Keith Jaggers & Ted Robert Gurr, 2006. Polity IV Project.

University of Maryland.

Mickolus, Edward, Todd Sandler & Jean Murdock, 1989. International Terrorism in the

1980s. Volume I 1980-1983. Ames: Iowa State University Press.

Mickolus, Edward, Todd Sandler, Jean Murdock & Peter Flemming, 2003. Interna-

tional Terrorism: Attributes of Terrorist Events (ITERATE). Data Codebook.

National Commission on Terrorist Attacks upon the United States, 2004. The 9/11

Commission Report. Washington, DC.

Neumayer, Eric & Thomas Plümper, 2009. International Terrorism and the Clash of

Civilizations. British Journal of Political Science, 39 (4): 711-734.

Office of the Coordinator for Counterterrorism, various years. Country Reports on

Terrorism. Washington, DC: US Department of State.

Pape, Robert A., 2003. ‘The Strategic Logic of Suicide Terrorism’, American Political

Science Review 97(3): 343-361.

Pape, Robert A., 2005. Dying to win – the strategic logic of suicide terrorism. New

York: Random House.

Piazza, James A., 2006. Rooted in Poverty? Terrorism, Poor Economic Development

and Social Cleavages. Terrorism and Political Violence 18(1): 159-177.

Pillar, Paul R., 2001. Terrorism and U.S. foreign policy. Washington, DC: Brookings

Institution Press.

38

Plümper, Thomas & Eric Neumayer, 2010. The Friend of my Enemy is my Enemy:

International Alliances and International Terrorism. European Journal of

Political Research, 49 (1): 75-96.

Plümper Thomas & Vera E. Troeger, 2007. ‘Efficient Estimation of Time-Invariant and

Rarely Changing Variables in Finite Sample Panel Analyses with Unit Fixed

Effects’, Political Analysis 15(2): 124-139.

Posen, Barry R., 2001/02. ‘The Struggle Against Terrorism – Grand Strategy, Strategy

and Tactics’, International Security 26(3): 39-55.

Powell, Robert, 2007. Defending against Terrorist Attacks with Limited Resources.

American Political Science Review 101(3): 527-541.

Rapoport, David C., 2004. ‘The Four Waves of Modern Terrorism’, in Audrey K.

Cronin & James M. Ludes, eds., Attacking Terrorism – Elements of a Grand

Strategy, Washington, DC: Georgetown University Press (46-73).

Richardson, Louise, 2006. What Terrorists Want. New York: Random House.

Rosendorff, B. Peter & Todd Sandler, 2004. ‘Too much of a good thing? The proactive

response dilemma’, Journal of Conflict Resolution 48(5): 657-671.

Sambanis, Nicholas, 2008. ‘Terrorism and civil war’, in Phil Keefer & Norman Loayza,

eds., Terrorism, Economic Development, and Political Openness. New York:

Cambridge University Press (174-206).

Shughart, William F., 2006. ‘An analytical history of terrorism, 1945-2000’, Public

Choice 128(1-2): 7-39.

SIPRI, 2007. Financial value of the arms trade.

(http://www.sipri.org/contents/armstrad/at_gov_ind_data.html).

39

Sobek, David & Alex Braithwaite, 2005. ‘Victim of success: American dominance and

terrorism’, Conflict Management and Peace Science 22(2): 135-148.

US Department of Defense, various years. Military Personnel Statistics

(http://siadapp.dior.whs.mil/personnel/MILITARY/Miltop.htm).

US Department of State, various years. World Military Expenditures and Arms

Transfers (http://www.state.gov/t/vci/rls/rpt/wmeat/ and

http://www.icpsr.umich.edu/).

USAID, 2006. U.S. Overseas Loans and Grants Online (Greenbook)

(http://qesdb.cdie.org/gbk/index.html).

Victoroff, Jeff, 2005. ‘The Mind of the Terrorist’, Journal of Conflict Resolution 49(1):

3-42.

World Bank, 2006. World Development Indicators on CD-Rom. Washington, DC:

World Bank.